REDRAFT scoring. Free rankings, our model projections. Smart tiers via GMM on (μ, σ).
Players are grouped into tiers by a Gaussian Mixture Model on each player's projection and projection uncertainty (the dispersion of their 3-/5-/10-game rolling averages). Tiers reflect where the distributions actually overlap, not arbitrary point cutoffs. Click any player card to mark them drafted.
Example projection range for the current QB rankings view.
Players are clustered with a heteroscedastic Gaussian Mixture Model. Each player has a projection (μ) and a derived uncertainty (σ) from the spread of their 3-, 5-, and 10-game rolling averages. The model groups players with overlapping distributions into the same tier, and BIC picks the number of tiers automatically. This is the same idea Boris Chen popularized for fantasy football rankings, adapted for per-player rolling-window data.
It is the ratio of a player’s posterior probability for their assigned tier vs. the next-best tier. ≥ 3 means tier lock, < 1.5 means tier bubble — a player who could plausibly drop or rise a tier.
Click any card to mark a player drafted; they will be struck through and persist across reloads on your device. Toggle to "Whole board" to see all positions side-by-side. Use the K slider to make tiers tighter or looser per position.
We use cookies for essential site functionality. With your consent, we also use cookies for analytics and performance monitoring. See our Privacy Policy.
